Abstract:
Internet has resulted an enormous rise in number of online social network (OSNs) users and popularity of social media. Privacy and security of OSNs require evaluation from different angles. Online rating systems are broadly acknowledged as resource for quality assessment on the web and users progressively rely on these platforms while they are deciding to do a transaction online. This makes such rating systems prone and common targets of attempted manipulation by posting unfair rating and reviews. Among millions of ratings generated online, no doubt many users post unfair response. Thus, finding these fake ratings is an essential task. To, overcome this issue an algorithm is proposed that calculates product reputation and users weights through ratings provided by the users. In this thesis we have analyzed already proposed methods and algorithms to detect fake ratings and address reputation issues in online social networks. For product reputation and data distribution plotting a statistical approach z-score is used and edges of violin graphs defines the outliers. The effectiveness of the algorithm is validated by conducting experiments in different scenarios on the datasets from MovieLens and Amazon. The algorithm using weighted averaging effectively highlighted the outliers.
